Generative AI Powers Geospatial Breakthroughs for Pipeline Leak Detection
By Sean Donegan, President and CEO, Satelytics
(P&GJ) — Over the last three years, the rise of generative AI has overhauled huge swaths of the economy. Everything from the writing of marketing copy to the process of new drug discovery has been transformed by large language models like ChatGPT.
These developments have, understandably, received a lot of attention. But they have arrived concurrently with another AI-related shift, one that does much the same thing for the actual, physical world: the rise of AI-powered geospatial analytics.
The primary use case here is the energy industry, which for decades now has been trying to reckon with the problem of its aging, leak-prone pipeline infrastructure. These companies want to be proactive – to plug these leaks before they can spiral out of control – but, given the size of the terrain, they are struggling even to locate these leaks in the first place.
AI-powered geospatial analytics is uncovering these leaks and placing the oil and gas industry on a much stronger footing when it comes to disaster prevention. The parallels with the ongoing GenAI revolution are – to say the least – striking.
Transforming Data at Scale
At the heart of GenAI is, of course, data. Large language models and generative AI are trained on inconceivably vast quantities of it; from these troves, they learn patterns, enabling them to coherently respond to a nearly infinite variety of prompts from users.
AI-powered geospatial analytics functions roughly the same way, while targeting a highly specialized problem set: the issues of leak and methane emission detection that plague oil and gas businesses.
Imagine the logistics involved in patrolling a given company's pipelines. As a rule, these pipelines are widely dispersed: For example, Plains All American Pipeline has a crude oil pipeline system of 18,300 miles (29,450 km). Properly patrolling a territory that vast is impossible: essentially, it is like trying to monitor the entire country. Businesses would need thousands of dedicated employees working around the clock to get a handle on even a modest percentage of it.
What this means is that, more often than not, leak detection and remediation is a reactive process: oil and gas producers intervene after things have started to go wrong. Much of the time this isn't a problem: the leak is small, localized, and can be quickly fixed. Often, though, the consequences are much more significant.
This is where AI-powered geospatial analytics come in. Here, the data takes the form of high-resolution satellite imagery: Multispectral and hyperspectral imagery is gathered from satellites, providing a kind of large-scale X-ray of a given oil and gas producer’s pipeline operations. As in GenAI, advanced algorithms are then deployed to make sense of this data. In the case of geospatial analytics, which means identifying problem points like methane emissions, liquid leaks, or right-of-way encroachments across incomprehensibly vast terrains.
At the heart of both processes, then, is a kind of high-tech simplification: breaking the world down into data and reshaping that data into insights that people can use.
Reshaping Decision-Making
GenAI has fundamentally changed decision-making for a number of use cases and processes, augmenting human responsibilities. Many hospitals, for instance, already deploy AI assistants, which work to prioritize care and optimize the working hours of nurses and doctors.
AI-powered geospatial analytics serve a similar function, allowing oil and gas producers to leverage their limited resources.
In the days before geospatial analytics, leak detection meant sending humans into remote areas to conduct periodic inspections. Many leaks, of course, wound up going undetected until they reached a point where a landowner or other third party noticed pools of liquid gathered on the ground.
AI-powered geospatial analytics have changed all that. Today, the AI can function as a kind of dispatcher. The technology knows where people are located and how far they are from a given leak site. It can even stipulate how long a given task might take to complete. Taking into the full scope of a given business's resources and challenges, it can allocate resources with ruthless efficiency, cutting down on wasted time while drastically reducing the potential risk of disasters.
To date, the potential of AI-powered geospatial analytics in energy infrastructure monitoring has received little attention next to big-ticket GenAI use-cases like disease diagnosis and customer service automation. And yet the transformative potential of this technology is immense.
It would be difficult to overstate the negative consequences of an out-of-control leak: there is the cost in environmental damage, there is the cost in population welfare, there is the cost in negative press and consequent reputational damage. And that's not even to mention the literal costs in the form of regulatory fines and civil lawsuits, tallying into the millions, and in certain infamous cases into the billions.
At an extraordinarily rapid clip, AI-powered geospatial analytics are reducing the risks of these outcomes for energy companies across the globe. It is a revolution every bit as exciting as what's happening with generative AI.
Author: Sean Donegan is the president and CEO of Satelytics. He brings over thirty years of technology and software development experience to the company. A dynamic leader, Sean’s career has been focused on building companies through creativity and innovation, recruiting highly effective teams to solve customers’ toughest challenges. Sean founded or owned four successful software companies, most recently Sean Allen, LLC, which was focused on predictive analytics in the oil & gas marketplace.